The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing local Python scripts (solve_captcha.py and solve_image_grid.py) to interface with external CAPTCHA solving APIs and vision AI models.\n- [EXTERNAL_DOWNLOADS]: Installation of well-known Python packages (playwright, anthropic, openai) and the download of browser binaries are necessary for the skill's automation capabilities.\n- [DATA_EXFILTRATION]: To solve challenges, the skill transmits site-specific data (sitekeys, page URLs) and visual data (screenshots of CAPTCHA images) to third-party services like 2captcha, CapMonster, or vision providers like Anthropic. This data transfer is the primary intended function of the skill.\n- [PROMPT_INJECTION]: The skill interacts with untrusted data from web pages (extraction of sitekeys and task descriptions) which is then passed to external APIs for processing.\n
Ingestion points: Sitekeys, task text, and CAPTCHA images are extracted from target websites via browser automation.\n
Boundary markers: The skill does not use specific boundary markers or isolation instructions when passing extracted task text to vision AI providers.\n
Capability inventory: The skill possesses the ability to execute subprocesses, perform network requests to solving APIs, and inject JavaScript into the browser context.\n
Sanitization: The skill uses json.dumps() to sanitize solved tokens before injecting them into the DOM, which prevents secondary injection vulnerabilities during the token placement phase.

captcha